Programming Languages Third Edition. Chapter 7 Basic Semantics

Transcription

1 Programming Languages Third Edition Chapter 7 Basic Semantics

2 Objectives Understand attributes, binding, and semantic functions Understand declarations, blocks, and scope Learn how to construct a symbol table Understand name resolution and overloading Understand allocation, lifetimes, and the environment Programming Languages, Third Edition 2

3 Objectives (cont d.) Work with variables and constants Learn how to handle aliases, dangling references, and garbage Perform an initial static semantic analysis of TinyAda Programming Languages, Third Edition 3

4 Introduction Syntax: what the language constructs look like Semantics: what the language constructs actually do Specifying semantics is more difficult than specifying syntax Several ways to specify semantics: Language reference manual Defining a translator Formal definition Programming Languages, Third Edition 4

5 Introduction (cont d.) Language reference manual: Most common way to specify semantics Provides clearer and more precise reference manuals Suffers from a lack of precision inherent in natural language descriptions May have omissions and ambiguities Programming Languages, Third Edition 5

6 Introduction (cont d.) Defining a translator: Questions about a language can be answered by experimentation Questions about program behavior cannot be answered in advance Bugs and machine dependencies in the translator may become part of the language semantics, possibly unintentionally May not be portable to all machines May not be generally available Programming Languages, Third Edition 6

7 Introduction (cont d.) Formal definition: Formal mathematical methods: precise, but are also complex and abstract Requires study to understand Denotational semantics: probably the best formal method for the description of the translation and execution of programs Describes semantics using a series of functions This course will use a hybrid of informal description with the simplified functions used in denotational descriptions Programming Languages, Third Edition 7

8 Attributes, Binding, and Semantic Functions Names (or identifiers): a fundamental abstraction mechanism used to denote language entities or constructs Fundamental step in describing semantics is to describe naming conventions for identifiers Most languages also include concepts of location and value Value: any storable quantities Location: place where value can be stored; usually a relative location Programming Languages, Third Edition 8

9 Attributes, Binding, and Semantic Functions (cont d.) Attributes: properties that determine the meaning of the name to which they are associated Example in C: Attributes for variables and constants include data type and value Example in C: Attributes include function, number, names and data type of parameters, return value data type, body of code to be executed Programming Languages, Third Edition 9

10 Attributes, Binding, and Semantic Functions (cont d.) Assignment statements associate attributes to names Example: Associates attribute value 2 to variable x Example in C++: Allocates memory (associates location to y) Associates value Programming Languages, Third Edition 10

11 Attributes, Binding, and Semantic Functions (cont d.) Binding: process of associating an attribute with a name Binding time: the time when an attribute is computed and bound to a name Two categories of binding: Static binding: occurs prior to execution Dynamic binding: occurs during execution Static attribute: an attribute that is bound statically Dynamic attribute: an attribute that is bound dynamically Programming Languages, Third Edition 11

12 Attributes, Binding, and Semantic Functions (cont d.) Languages differ substantially in which attributes are bound statically or dynamically Functional languages tend to have more dynamic binding than imperative languages Static attributes can be bound during translation, during linking, or during loading of the program Dynamic attributes can be bound at different times during execution, such as entry or exit from a procedure or from the program Programming Languages, Third Edition 12

13 Attributes, Binding, and Semantic Functions (cont d.) Some attributes are bound prior to translation time Predefined identifiers: specified by the language definition Values true/false bound to data type Boolean maxint specified by language definition and implementation All binding times except execution time are static binding Programming Languages, Third Edition 13

14 Attributes, Binding, and Semantic Functions (cont d.) A translator creates a data structure to maintain bindings Can be thought of as a function that expresses the binding of attributes to names Symbol table: a function from names to attributes Programming Languages, Third Edition 14

15 Attributes, Binding, and Semantic Functions (cont d.) Parsing phase of translation includes three types of analysis: Lexical analysis: determines whether a string of characters represents a token Syntax analysis: determines whether a sequence of tokens represents a phrase in the context-free grammar Static semantic analysis: establishes attributes of names in declarations and ensures that the use of these names conforms to their declared attributes During execution, attributes are also maintained Programming Languages, Third Edition 15

16 Attributes, Binding, and Semantic Functions (cont d.) Programming Languages, Third Edition 16

17 Declarations, Blocks, and Scope Bindings can be implicit or explicit Example: Data type is bound explicitly; location of x is bound implicitly Entire declaration itself may be implicit in languages where simply using the variable name causes it to be declared Definition: in C and C++, a declaration that binds all potential attributes Prototype: function declaration that specifies the data type but not the code to implement it Programming Languages, Third Edition 17

18 Declarations, Blocks, and Scope (cont d.) Block: a sequence of declarations followed by a sequence of statements Compound statements: blocks in C that appear as the body of functions or anywhere an ordinary program statement could appear Local declarations: associated with a block Nonlocal declarations: associated with surrounding blocks Block-structured languages allow nesting of blocks and redeclaration of names within nested blocks Programming Languages, Third Edition 18

19 Declarations, Blocks, and Scope (cont d.) Each declared name has a lexical address containing a level number and an offset Level number starts at 0 and increases into each nested block Other sources of declarations include: A struct definition composed of local (member) declarations A class in object-oriented languages Declarations can be collected into packages (Ada), modules (ML, Haskell, Python), and namespaces (C++) Programming Languages, Third Edition 19

20 Declarations, Blocks, and Scope (cont d.) Scope of a binding: region of the program over which the binding is maintained Lexical scope: in block-structured languages, scope is limited to the block in which its associated declaration appears (and other blocks contained within it) Declaration before use rule: in C, scope of a declaration extends from the point of declaration to the end of the block in which it is located Programming Languages, Third Edition 20

21 Programming Languages, Third Edition 21

22 Programming Languages, Third Edition 22

23 Declarations, Blocks, and Scope (cont d.) Declarations in nested blocks take precedence over previous declarations A global variable is said to have a scope hole in a block containing a local declaration with the same name Use scope resolution operator :: in C++ to access the global variable Local declaration is said to shadow its global declaration Visibility: includes only regions where the bindings of a declaration apply Programming Languages, Third Edition 23

24 Declarations, Blocks, and Scope (cont d.) Programming Languages, Third Edition 24

25 Declarations, Blocks, and Scope (cont d.) Scope rules need to be constructed such that recursive (self-referential) declarations are possible when they make sense Example: functions must be allowed to be recursive, so function name must have scope beginning before the block of the function body Programming Languages, Third Edition 25

26 The Symbol Table Symbol table: Must support insertion, lookup, and deletion of names with associated attributes, representing bindings in declarations A lexically scoped, block-structured language requires a stack-like data structure to perform scope analysis: On block entry, all declarations of that block are processed and bindings added to symbol table On block exit, bindings are removed, restoring any previous bindings that may have existed Programming Languages, Third Edition 26

27 Programming Languages, Third Edition 27

28 The Symbol Table (cont d.) Programming Languages, Third Edition 28

29 The Symbol Table (cont d.) Programming Languages, Third Edition 29

30 The Symbol Table (cont d.) Programming Languages, Third Edition 30

31 The Symbol Table (cont d.) Programming Languages, Third Edition 31

32 The Symbol Table (cont d.) Programming Languages, Third Edition 32

33 The Symbol Table (cont d.) Programming Languages, Third Edition 33

34 Programming Languages, Third Edition 34

35 The Symbol Table (cont d.) The previous example assumes that declarations are processed statically (prior to execution) This is called static scoping or lexical scoping Symbol table is managed by a compiler Bindings of declarations are all static If symbol table is managed dynamically (during execution), declarations will be processed as they are encountered along an execution path This is called dynamic scoping Programming Languages, Third Edition 35

36 Programming Languages, Third Edition 36

37 Programming Languages, Third Edition 37

38 Programming Languages, Third Edition 38

39 Programming Languages, Third Edition 39

40 The Symbol Table (cont d.) Dynamic scoping will affect the semantics of the program and produce different output Output using lexical scoping: Output using dynamic scoping: Dynamic scope can be problematic, which is why few languages use it Programming Languages, Third Edition 40

41 The Symbol Table (cont d.) Problems with dynamic scoping: The declaration of a nonlocal name cannot be determined by simply reading the program: the program must be executed to know the execution path Since nonlocal variable references cannot be predicted prior to execution, neither can their data types Dynamic scoping is a possible option for highly dynamic, interpreted languages when programs are not expected to be extremely large Programming Languages, Third Edition 41

42 The Symbol Table (cont d.) Runtime environment is simpler with dynamic scoping in an interpreter APL, Snobol, Perl, and early dialects of Lisp were dynamically scoped Scheme and Common Lisp use static scoping There is additional complexity for symbol tables struct declaration must contain further declarations of the data fields within it Those fields must be accessible using dot member notation whenever the struct variable is in scope Programming Languages, Third Edition 42

43 The Symbol Table (cont d.) Two implications for struct variables: A struct declaration actually contains a local symbol table itself as an attribute This local symbol table cannot be deleted until the struct variable itself is deleted from the global symbol table of the program Programming Languages, Third Edition 43

44 Programming Languages, Third Edition 44

45 Programming Languages, Third Edition 45

46 The Symbol Table (cont d.) Any scoping structure that can be referenced directly must also have its own symbol table Examples: Named scopes in Ada Classes, structs, and namespaces in C++ Classes and packages in Java Typically, there will be a table for each scope in a stack of symbol tables When a reference to a name occurs, a search begins in the current table and continues to the next table if not found, and so on Programming Languages, Third Edition 46

47 Programming Languages, Third Edition 47

48 Programming Languages, Third Edition 48

49 Name Resolution and Overloading Addition operator + actually indicates at least two different operations: integer addition and floatingpoint addition + operator is said to be overloaded Translator must look at the data type of each operand to determine which operation is indicated Overload resolution: process of choosing a unique function among many with the same name Lookup operation of a symbol table must search on name plus number and data type of parameters Programming Languages, Third Edition 49

50 Name Resolution and Overloading (cont d.) Programming Languages, Third Edition 50

51 Name Resolution and Overloading (cont d.) Consider these function calls: Symbol table can determine the appropriate function based on number and type of parameters Calling context: the information contained in each call But this ambiguous call depends on the language rules (if any) for converting between data types: Programming Languages, Third Edition 51

52 Name Resolution and Overloading (cont d.) Adding these definitions makes the function calls legal in C++ and Ada but is unnecessary in Java Automatic conversions as they exist in C++ and Java significantly complicate overload resolution Programming Languages, Third Edition 52

53 Name Resolution and Overloading (cont d.) Additional information in a calling context may be used for overload resolution: Ada allows the return type and names of parameters to be used for overhead resolution C++ and Java ignore the return type Both Ada and C++ (but not Java) allow built-in operators to be overloaded When overloading a built-in operator, we must accept its syntactic properties Example: cannot change the associativity or precedence of the + operator Programming Languages, Third Edition 53

54 Name Resolution and Overloading (cont d.) Note that there is no semantic difference between operators and functions, only syntactic difference Operators are written in infix form Function calls are always written in prefix form Names can also be overloaded Some languages use different symbol tables for each of the major kinds of definitions to allow the same name for a type, a function, and a variable Example: Java Programming Languages, Third Edition 54

55 Name Resolution and Overloading (cont d.) Programming Languages, Third Edition 55

56 Allocation, Lifetimes, and the Environment Environment: maintains the bindings of names to locations May be constructed statically (at load time), dynamically (at execution time), or with a mixture of both Not all names in a program are bound to locations Examples: names of constants and data types may represent purely compile-time quantities Declarations are also used in environment construction Indicate what allocation code must be generated Programming Languages, Third Edition 56

57 Allocation, Lifetimes, and the Environment (cont d.) Typically, in a block-structured language: Global variables are allocated statically Local variables are allocated dynamically when the block is entered When a block is entered, memory for variables declared in that block is allocated When a block is exited, this memory is deallocated Programming Languages, Third Edition 57

58 Programming Languages, Third Edition 58

59 Allocation, Lifetimes, and the Environment (cont d.) Programming Languages, Third Edition 59

60 Allocation, Lifetimes, and the Environment (cont d.) Programming Languages, Third Edition 60

61 Allocation, Lifetimes, and the Environment (cont d.) Programming Languages, Third Edition 61

62 Allocation, Lifetimes, and the Environment (cont d.) Programming Languages, Third Edition 62

63 Allocation, Lifetimes, and the Environment (cont d.) Memory for local variables within a function will not be allocated until the function is called Activation: a call to a function Activation record: the corresponding region of allocated memory In a block-structured language with lexical scope, the same name can be associated with different locations, but only one of these can be accessed at any one time Lifetime (or extent) of an object is the duration of its allocation in the environment Programming Languages, Third Edition 63

64 Allocation, Lifetimes, and the Environment (cont d.) Lifetime of an object can extend beyond the region of a program in which it can be accessed Lifetime extends through a scope hole Pointer: an object whose stored value is a reference to another object C allows the initialization of pointers that do not point to an allocated object: Objects must be manually allocated by use of an allocation routine Variable can be dereferenced using the unary * operator Programming Languages, Third Edition 64

65 Allocation, Lifetimes, and the Environment (cont d.) C++ simplifies dynamic allocation with operators new and delete: These are used as unary operators, not functions Heap: area in memory from which locations can be allocated in response to calls to new Dynamic allocation: allocation on the heap Programming Languages, Third Edition 65

66 Programming Languages, Third Edition 66

67 Allocation, Lifetimes, and the Environment (cont d.) Many languages require that heap deallocation be managed automatically Heap allocation/deallocation and explicit pointer manipulation are inherently unsafe operations Can introduce seriously faulty runtime behavior that may even compromise the operating system Storage class: the type of allocation Static (for global variables) Automatic (for local variables) Dynamic (for heap allocation) Programming Languages, Third Edition 67

68 Variables and Constants Although references to variables and constants look the same in many languages, their roles and semantics are very different We will look at the basic semantics of both Programming Languages, Third Edition 68

69 Variables Variable: an object whose stored value can change during execution Is completely specified by its attributes (name, location, value, data type, size of memory storage) Box and circle diagram: focuses on name and location Programming Languages, Third Edition 69

70 Variables (cont d.) Assignment statement: principle way in which a variable changes its value Example: Semantics: expression e is evaluated to a value, then copied into the location of x If e is a variable named y: Programming Languages, Third Edition 70

71 Variables (cont d.) Variable on right side of assignment statement stands for a value (r-value); variable on left side stands for a location (l-value) Address of operator (&) in C: turns a reference into a pointer to fetch the address of a variable Assignment by sharing: the location is copied instead of the value Assignment by cloning: allocates new location, copies value, and binds to the new location Both are sometimes called pointer semantics or reference semantics Programming Languages, Third Edition 71

72 Variables (cont d.) Programming Languages, Third Edition 72

73 Variables (cont d.) Storage semantics or value semantics refer to standard assignment Standard implementation of assignment by sharing uses pointers and implicit dereferencing Programming Languages, Third Edition 73

74 Variables (cont d.) Programming Languages, Third Edition 74

75 Constants Constant: an entity with a fixed value for the duration of its existence in a program Like a variable, but has no location attribute Sometimes say that a constant has value semantics Literal: a representation of characters or digits Compile-time constant: its value can be computed during compilation Static constant: its value can be computed at load time Programming Languages, Third Edition 75

76 Constants (cont d.) Manifest constant: a name for a literal Dynamic constant: its value must be computed during execution Function definitions in virtually all languages are definitions of constants whose values are functions This differs from a function variable in C, which must be defined as a pointer Programming Languages, Third Edition 76

77 a and b are compile-time constants a is a manifest constant c is a static (loadtime constant) d is a dynamic constant Constants (cont d.) Programming Languages, Third Edition 77

78 Aliases, Dangling References, and Garbage There are several problems with naming and dynamic allocation conventions of programming languages, especially C, C++, and Ada As a programmer, you can learn to avoid those problematic situations As a language designer, you can build solutions into your language Programming Languages, Third Edition 78

79 Aliases Alias: occurs when the same object is bound to two different names at the same time Can occur during procedure call, through the use of pointer variables, or through assignment by sharing Programming Languages, Third Edition 79

80 Aliases (cont d.) Programming Languages, Third Edition 80

81 Aliases (cont d.) Programming Languages, Third Edition 81

82 Aliases (cont d.) Programming Languages, Third Edition 82

83 Aliases (cont d.) Aliases can potentially cause harmful side effects Side effect: any change in a variable s value that persists beyond the execution of the statement Not all side effects are harmful; an assignment statement is intended to cause one Side effects that change variables whose names do not directly appear in the statement are potentially harmful Cannot be determined from the written code Aliasing due to pointer assignment is difficult to control Programming Languages, Third Edition 83

84 Aliases (cont d.) Assignment by sharing implicitly uses pointers Java has a mechanism for explicitly cloning an object so that aliases are not created by assignment Programming Languages, Third Edition 84

85 Dangling References Dangling reference: a location that has been deallocated from the environment but can still be accessed by a program Occurs when a pointer points to a deallocated object Programming Languages, Third Edition 85

86 Dangling References (cont d.) Can also result from automatic deallocation of local variables on exit from a block, with the C address of operator Programming Languages, Third Edition 86

87 Dangling References (cont d.) Java does not allow dangling references at all because: There are no explicit pointers There is no address of operator There are no memory deallocation operators such as free or delete Programming Languages, Third Edition 87

88 Garbage Garbage: memory that has been allocated in the environment but is now inaccessible to the program Can occur in C by failing to call free before reassigning a pointer variable A program that is internally correct but produces garbage may run out of memory Programming Languages, Third Edition 88

89 Garbage (cont d.) A program with dangling references may: Produce incorrect results Corrupt other programs in memory Cause runtime errors that are hard to locate For this reason, it is useful to remove the need to deallocate memory explicitly from the programmer Garbage collection: process of automatically reclaiming garbage Language design is a key factor in the kind of runtime environment necessary for correct execution of programs Programming Languages, Third Edition 89

90 Case Study: Initial Static Semantic Analysis of TinyAda Chapter 6 introduced a syntax analyzer for TinyAda A simple parsing shell that pulled tokens from a scanner until a syntax error was detected Here, we extend the parsing shell to perform some semantic analysis Focus will be on tools for scope analysis and for restricting the use of identifiers Must focus on two attributes of an identifier: Name Role it plays (constant, variable, type, or procedure) Programming Languages, Third Edition 90

91 Scope Analysis TinyAda is lexically scoped, with these scope rules: All identifiers must be declared before use At most, one declaration for a given identifier in a single block A new block starts with formal parameter specifications of a procedure and extends to the reserved word end Visibility of a declared identifier extends into nested blocks unless it is redeclared in that block Identifiers are not case sensitive Programming Languages, Third Edition 91

92 Scope Analysis (cont d.) TinyAda has five built-in (predefined) identifiers: Data type names integer, char, boolean Boolean constants true and false These identifiers must be visible in a top-level scope before a source program is parsed Static nesting level of this scope is 0 Scope at nesting level 1 contains the procedure s formal parameters (if any) and any identifiers introduced in the procedures basic declarations Names in nested procedures follow this pattern Programming Languages, Third Edition 92

93 Scope Analysis (cont d.) TinyAda s parser uses a stack of symbol tables When each new scope is entered, a new table is pushed onto the stack When a scope is exited, the table at the top of the stack is popped off the stack Two classes are defined to support scope analysis: SymbolEntry: holds information about an identifier SymbolTable: manages the stack of scopes Programming Languages, Third Edition 93

94 Scope Analysis (cont d.) Programming Languages, Third Edition 94

95 Identifier Role Analysis An identifier names an entity, such as a variable, a constant, or an entire data type This attribute of an identifier is called its role An identifier s role imposes certain restrictions on its use Examples: Only a variable or parameter identifier can appear on the left side of an assignment statement Only a type identifier can appear as the element type of an array Programming Languages, Third Edition 95

96 Identifier Role Analysis (cont d.) Identifier acquires its role in its declaration Role is saved in the symbol table for future use Role analysis uses the symbol table to share contextual information about identifiers among otherwise independent parsing methods Programming Languages, Third Edition 96